(Page 1 of 5 pages for this article  1 2 3 >  Last »)

Filed under:

Searchable Video with Creative Suite 4

With the proliferation of video on the Web, access to information is getting easier to find and understand (for most).  The challenge is twofold: first is getting people to find and search your video for the right information and second is if the individual has an auditory impairment, so much information is only available in the audio track.  This is why there has been such a big push to make video searchable online and more accessible. In this article you’ll use Adobe’s tool set to set up your video content for both goals.

There are two reasons to give everyone equal access to the information. The first is simple, accessible video can be highly-optimized for search engines.  This means its easier for people to find, and even search, your video. Video is one of the most expensive type of web content.  By getting more eyeballs on it, you’re improving the chances of recouping the investment.

The second reason?  It’s the right thing to do (and in some cases its even the law).  You need to do whatever you can to make video accessible to as many people as possible. We’ll show you several things that can be done to greatly improve accessibility.

To make searchable video, you’ll use Adobe Soundbooth CS4 to transcribe a video’s audio track and then you’ll turn that transcription into Flash-video cue points.  Once that’s done you’ll learn how to make video searchable and how to display captions while the video is playing. Together, this makes your video truly accessible.

This tutorial is adapted from the new Adobe Press book After Effects for Flash | Flash for After Effects (co-written by Marcus Geduld). You will need these files and source code for the tutorial if you want to try this out.  It’s a free download (no login required).

To whet your appetite, you can click here to see the end result. (Will open in a new window)

Transcribing Audio with Soundbooth

This first step towards searchable video is having a transcript. Adobe Soundbooth offers a very useful option, called Speech Search, that recognizes speech and translates it into text. Soundbooth is included as a standalone product as well as in the Production Premium and Master Collection CS4 bundles.

The accuracy of this transcription can teeter-totter between being incredibly accurate to not very accurate. There are a few factors you can influence to improve the accuracy of the text:

• Use a professional-quality microphone to get a cleaner audio signal.
• Try to minimize or remove background noise.
• Isolate the voice from music or sound effects to make the transcription more accurate.
• Encourage the speaker to clearly pronounce words.

You will find that you’ll generally have to do some cleanup after the automatic transcription process. Fortunately, Soundbooth makes cleanup easy.

1. Copy the Chapter_14 Project Files folder from the DVD to a local hard drive. This is the free download that you got here.

2. Launch Soundbooth, and then choose Window > Workspace > Edit Audio to Video to place Soundbooth in a state to work with video files.

3. Choose Window > Metadata to view the Metadata panel. The Metadata panel is your starting point for speech transcription. Running the transcription function takes a while, but when it’s finished, all metadata for clips will be searchable from the Metadata panel. Fortunately, the process can run in the background, which means you can still edit or perform other tasks. The Metadata panel also allows you to search by any word or phrase within the clip. 

4. Choose File > Open and open the file gasPrices_original.f4v from the Chapter_14 Project Files folder. You can open any standard type of video file in Soundbooth, such as QuickTime, AVI, F4V, or FLV.

5. Click Play to preview the file. Become familiar with the short news story. You will optimize this clip for the Web and make it more accessible for the news channel’s Web site.

(Page 1 of 5 pages for this article  1 2 3 >  Last »)

Amazing. I hope FCS3 has this feature. Would be killer for the doc world (and as the tech continues to evolve, even more so) in addition to online search and other applications. Thanks!

Posted by Jared Scheib  on  06/17  at  01:39 AM

---

I’m having problems with the first step. I tried to load an audio track in Soundbooth but the results says error in loading. It’s in wma by the way, do I need to convert it?

Posted by diesel generators  on  08/18  at  03:25 PM

---

By processing RM if you mean RealMedia… no.  Convert to MP3 or AIFF. Less compression actually translates cleaner.

Posted by Richard Harrington  on  08/22  at  02:33 PM

---

diesel…

Yes you need to convert files.

I suggest you try the demo files we give you first.

Posted by Richard Harrington  on  08/22  at  02:34 PM

---

Be sure that you use the right file formats… not all file types are supported.

Posted by Richard Harrington  on  12/05  at  03:13 PM

---

Great artical! Is there an .FLA file example?

Posted by .(JavaScript must be enabled to view this email address)  on  12/10  at  02:45 PM

---

Hi,
I came to this demo rather late as I’m using Sb CS5, and I was rather disappointed to find that this searchable-video-prog doesn’t work when it’s prepared in the CS5 version. I’m not sure, but it may be that Sb CS5 does not export exactly the same data in its xml file(?)  All the speech analysis, exporting, re-importing as markers, and then saving of the vid+metadata… all work well, but when launched in the browser, it stops at the “Please be patient… etc” dialogue box and never gets any further.

What would be really good, would be to have a copy of the .fla file of the video player, with its ActionScript, so I could have a go at matching the code that reads in the data with those in the exported xml file. Do you think there is any chance of that happening?

Posted by .(JavaScript must be enabled to view this email address)  on  09/01  at  02:53 AM

---

Name:

Email:

Location:

URL:

Smileys

Remember my personal information

Notify me of follow-up comments?

Submit the word you see below: